JPH06266650A - Method and equipment for transferring data and interleaving device for data transfer - Google Patents

Abstract

PURPOSE: To interleave data transfer for each slice by dividing the data transfer into the sequence of plural data slices by each channel. CONSTITUTION: In plural DMA channels, at the time of transferring the first number of data through a first DMA channel, when the first number is less than a prescribed channel interleave size, the data are divided into plural slices and transferred. After the slice of the data in the first DMA channel, inquiry is operated for judging whether or not a second DMA channel obtains the control of a transfer resource, and when the second DMA channel does not obtain the control, the slices of the data are continuously transferred through the first DMA charnel. Also, the control is moved to the second DMA channel so that the second DMA channel can transfer the data, and the inquiry can be operated after each slice.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to data transfer operations in computer systems, and more particularly to direct memory access operations in computer systems.

[0002]

2. Description of the Related Art In a computer system having a plurality of data processing devices, not only the transfer of data between one device and another device but also one device and another device must be performed in order to obtain system efficiency. It also happens in a time-multiplexed manner with other devices. For example, the central processing unit (“CP
U ') would be able to transfer data to one network device while time-multiplexing the data transfer activity with other devices to fully utilize the CPU. Since the data transfer rates from various devices may be different, a buffer may be provided as an intermediary between each device and the CPU to optimize the data transfer rates of multiple devices. If a buffer is provided, until the transfer resource becomes available or when the transfer destination is available,
Data can be stored temporarily. Providing a buffer for each device presents a problem for integrated circuits as more sophisticated systems are often required as more devices interact with the CPU. In integrated circuits,
Due to die size limitations, more buffers and associated decoding logic cannot be added.

This problem can be illustrated by a direct memory access operation that transfers data through multiple channels in a computer system. Direct memory access (“DMA”) operations are a technique used for computer input / output (“I / O”) operations when large amounts of data are to be moved. DMAs typically include additional modules on the system bus. The DMA module 100 as shown in FIG. 1 can emulate a CPU (not shown) and take over control of the system bus from the CPU. The DMA operation proceeds as follows. CPU is
When it wants to read or write a block of data, it issues a command to the DMA module 100. The command includes information as to whether a read 101 was requested or a write 102 was requested, the address 103 of the associated I / O device, the memory location where the read or write should begin, the word to read or Number of words to write 1
Including 05 and. Therefore, the CPU DMs this I / O operation.
Since it is entrusted to the A module, it continues to execute other tasks, and the module processes the task. Therefore, the DMA module transfers all the data blocks directly to and from the memory, one word at a time, without passing through the CPU. When the transfer is complete, the DMA module sends an interrupt signal 110 to the CPU. Therefore, the CPU need only participate in the operation at the beginning and end of the transfer.

The DMA module needs to take over control of the bus in order to transfer data to and from memory. For this purpose, the DMA module must use the bus only when the CPU does not need it, or the DMA module must cause the CPU to suspend its operation. This DM
Since the only function performed by the A module is data transfer, the transfer sequence can be hardwired into the module's circuitry. By fetching instructions at a much higher level, bandwidth utilization is minimized. Because the DMA module has the ability to generate the address and control signals required by the bus, the DMA module can perform I / O operations at full memory speeds.

In today's efficient computer systems, DMA operations must also accommodate multi-channel data transfers to and from various devices. A separate buffer is assigned to the DMA channel to facilitate data transfer. However, the operation became more elaborate and DMA
As the number of channels increases, the "one buffer per channel" scheme is not entirely impractical,
Is no longer desirable. Furthermore, first-in first-out ("F
A single buffer, such as an IFO ") buffer, may serve as an intermediary for multi-channel DMA operations,
FIFO if one channel that needs all the data immediately is blocked by the data of another channel
But the problem of bottleneck still occurs. Also, F
The FIFO is constrained to fill while data is being transferred to the IFO, forcing the transfer to stop until some data in the FIFO is read. For the above reasons, the conventional FIFO in the multi-channel transfer cannot execute the continuous operation having the wide bandwidth.

As will be described below, the present invention provides for multiple bandwidths over multiple DMA channels by interleaving the transfer sequences from different channels in order to obtain wide bandwidth and efficiently utilize resources. A method and apparatus for transferring data is disclosed. Also disclosed is a circuit architecture that is most advantageous for use in combination with the support of multiple channel DMA transfers to maximize bandwidth and system efficiency.

[0007]

Accordingly, it is an object of the present invention to support multi-channel DMA operation. It is also an object of the present invention to support multiple channel DMA transfers in a continuous operational flow. Another object of the invention is to achieve high bandwidth multi-channel DMA transfers by interleaving the channels. Another object of the present invention is to provide a circuit architecture that is most advantageous for use in combination with multiple interleaved DMA channels. Another object of the present invention is to provide a plurality of D's while independently reading and writing for each channel.
It is to support the circuit architecture that supports the MA channel. Another object of the invention is to interleave D
It is to provide a circuit architecture for efficiently using a bus for MA channels.

[0008]

As described below, the present invention supports multiple channel DMA operations by dividing the data transfer for each channel into a sequence of data slices and interleaving the transfers slice by slice. A method and apparatus for doing so is disclosed. Control of transfer resources may be transferred between DMA channels, but the ordering of data slices per channel is maintained. Further, the present invention discloses a circuit architecture that is most advantageous for use with multiple interleaved DMA channels. This circuit architecture comprises a dual port memory, a channel sequencer and a channel interleave controller. The dual port memory stores a slice of data to be transferred over the channel. The channel sequencer maintains the channel ordering of the data slices in dual port memory. A channel interleaver causes a channel to interleave its data transfers by monitoring the channel interleave size, the current data transfer count, and the total transfer count for each channel. The second channel transmits data over the same medium as the first channel when the first channel reaches its channel interleave size or when the first channel has finished transmitting the requested total transfer count. Are allowed to be transferred, which allows efficient use of the bus. Objects, features and advantages of the present invention will be apparent from the following detailed description.

Notation and Terminology Most of the following detailed description is presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. Such algorithmic representations and representations are the means used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. Here, and generally, an algorithm is considered to be a consistent sequence of steps leading to a desired result. The steps are those requiring physical manipulations of physical quantities. Usually, those quantities take the form of electrical or magnetic signals that can be stored, transferred, combined, compared and otherwise manipulated, but not necessarily. At times, it has proved convenient to refer to such signals as bits, values, elements, symbols, characters, terms, numbers, etc., primarily because of their commonly used terminology. It should be remembered, however, that all such terms and related terms should be associated with the appropriate physical quantities and are merely labels attached to those quantities for convenience.

Further, although the operations performed are often referred to in terms, such as adding or comparing, which are commonly associated with intelligent operations performed by an operator, the operations described herein are part of the present invention. Whichever of the above, the ability of such an operator is unnecessary and is often undesirable.
The motion is the motion of the machine. Useful machines for performing the operations of the present invention include general purpose digital computers or other similar devices. In each case, attention should be paid to the clear distinction between the method operations when operating a computer and the method of computation itself. The present invention relates to method steps for operating a computer in processing electrical signals or other physical (eg, mechanical, chemical) signals to generate another desired physical signal. The invention also relates to a device for performing such an operation. This device may be specially constructed for the required purposes, or a general purpose computer may be selectively activated or reconfigured by a computer program stored in the computer. The algorithm presented here is originally
It is not associated with any particular computer or other particular apparatus. That is, it will be appreciated that a variety of general purpose machines may be used with the programs written in accordance with the teachings of the present invention, or it may be more convenient to construct a more specialized apparatus to perform the required method steps. It may be. The required structure for a variety of such machines will appear from the description below.

Coding Details No particular programming language was dictated in the performance of the various procedures described herein. This is because, in part, not all languages that could be mentioned are universally available. Each user of a particular computer is aware of the language that best suits its immediate purpose. In practice, it has been found useful to substantially implement the invention with an assembly language that provides machine-executable object code. A computer or monitor system that can be used to implement the present invention is made up of many various elements, and therefore a detailed program list is not given. The operations and other procedures described herein and illustrated in the accompanying drawings are considered to be sufficiently disclosed to enable one skilled in the art to practice the invention.

[0012]

DETAILED DESCRIPTION As will be appreciated by those skilled in the art, the channel interleaving method and apparatus of the present invention divides a data transfer sequence into a plurality of slices and divides those slices from a plurality of different channels. By interleaving, multi-channel transfer can proceed in an orderly and efficient manner. The present invention further discloses a circuit architecture that supports interleaving channels so that sliced data is transferred over respective corresponding channels. That is, the channel interleaving method and the channel interleaving apparatus more efficiently use the bus and extend the bandwidth by interleaving the other channel when one channel does not use the transfer resource.
In the following description, for the sake of convenience, the specific memory, organization, architecture, etc. are set forth in order to provide a thorough understanding of the present invention, but it should be understood that the present invention may be practiced without the specific details. It will be obvious to the trader. Also, in some instances, well-known circuits may be shown in block diagram form in order not to unnecessarily obscure the present invention.
Moreover, as will be apparent from the following description, the channel interleaving method and apparatus may utilize other buffering schemes other than the unique circuit architecture described herein. Similarly, unique circuit architectures may be utilized in combination with other inter-device transfers in addition to the inventive interleaving method and interleaving device disclosed herein.

Referring to FIG. 2, a computer system with a DMA controller is shown in a block diagram.
The DMA controller module 220 facilitates data transfer between the memory 240 and various peripheral devices 250 by gaining control of the system bus 200 when the CPU 210 is not using the system bus 200. A buffer 230 coupled to the DMA controller module 220 performs data buffering during data transfer. As will be appreciated by those skilled in the art, an architecture capable of supporting multiple DMA channels significantly improves the performance of DMA operations between memory 240 and various peripherals 250.

Referring now to FIG. 7, there is shown a flowchart of a DMA channel interleaving method. First, the total transfer count of data (total_transfer_count, “T
Issue a DMA request to transfer TC "). Determine whether a transfer sequence should be split into two or more separate slices so that a slice from channel M can be interleaved with a slice from another channel N wishing to use the same transfer resource. In order to set the TTC to a predetermined channel interleave size (ch
annel_interleave_size, “CI
S "). TTC from channel M is its CI
If it is S or less, the transfer is completed in one slice. Incidentally, D
The CIS of an MA channel specifies the amount of data that a DMA channel must transfer before another DMA channel can interleave using the same transfer resources as the first DMA channel. Therefore, when the TTC of the channel M is within the CIS, the data can be easily transferred by one slice.

If the CIS of channel M is smaller than the required TTC, the TTC transfer is divided into a plurality of CIS size data slices according to time. Each slice can be transferred via transfer resources. After the transfer of each slice for channel M is complete, the transfer resource control is available for interleaving another slice of data from another channel. The channel priority arbitration scheme allows control of transfer resources to be transferred to another channel of higher priority.
The transfer may be a rotary system or a fixed system. Channel M slice followed by another channel N
When the control resource of the transfer resource is acquired, the data slice from the channel N can be transferred using the transfer resource. After the end of each slice, control of transfer resources is further transferred from channel N to another channel of higher priority (if such a channel exists). Current transfer count (current_transfer) of each channel
It will be apparent to those skilled in the art that a counting mechanism should be provided to keep track of (_count, "CTC"). Therefore, when data transfer is performed through the same transfer resource for a plurality of DMA channels, the data transfer sequence through each DMA channel is divided into a plurality of slices according to the TTC and CIS according to time,
Interleave with other DMA channels. As a result, multiple DMA data transfers are channel interleaved in random order by slice until the entire sequence is complete.

Although various buffering embodiments of the present invention capable of supporting channel interleaving can be configured by those skilled in the art, one of the circuit architectures supporting channel interleaving will be described below with reference to FIG. This will be described as a preferred embodiment. The control structure 300 comprises a dual port memory 320 and a channel sequencer 330. The dual port memory 320 writes and reads data for the DMA data transfer and the read pointer 3
21 and the write pointer 325 operate in a circular manner. The write pointer 325 points to the data storage location of the dual port memory 320 to be written next. When writing data to the dual port memory 320,
The write pointer 325 moves to the next storage location in the dual port memory 320. Read pointer 321 points to the data storage location of dual port memory 320 to be read next. Therefore, the dual port memory 320
The data is read from the storage location designated by the read pointer 321 of. The write pointer 325 is the dual port memory 320 when the two pointers are coincident with each other, ie, “bump up”, and the dual port memory 320 is empty when the write pointer 325 is located immediately after the read pointer 321. Is a leading pointer that is full. Usually, control structure 3
00 starts operation in an empty state. Further, those skilled in the art will appreciate that the write pointer 325 should never overtake the read pointer 321 so that valid data is not overwritten before it is read from the dual port memory 320. Let's do it.

Referring to FIG. 3, since the channel sequencer 330 records the DMA channel number associated with each DMA data transfer through the dual port memory 320, a plurality of channels are DMA with the dual port memory 320. When performing a transfer, the channel ordering corresponding to the data buffered in dual port memory 320 is maintained. The channel sequencer 330 has a source pointer 331 and a destination pointer 335.
Operates in a circulation system with and. Source pointer 331
Writes a channel number, which represents the channel providing the data to be written, to one memory location of the channel sequencer 330. The source pointer 331 is the channel sequencer 330
Is incremented to the next memory location in the memory and ready for the next channel number to be written. The destination pointer 335 directs one storage location of the channel sequencer 330 to read data from the dual port memory 320 to the channel pointed to by the destination pointer 335 in the channel sequencer 330, thereby causing the dual port memory 3 to operate.
The data transfer from 20 is started. In this way, the writing of data to the dual port memory 320 and the reading of data from the dual port memory 320 are performed in the same channel. It should be apparent to those skilled in the art that the source pointer 331 and the destination pointer 335 do not cross each other to avoid misleading the transfer of data.

Further description will be continued with reference to FIG. If a resource such as a bus is not available to facilitate data transfer, the source pointer 331 skips its storage location and tags it with the tag 332 of the channel sequencer 330. So, when the destination pointer 335 moves to the tagged storage location, the destination pointer 335 reads the tag 332 and that the source pointer 331 has not yet performed a data transfer for the tagged storage location of the channel sequencer 330. To know As a result, the destination pointer 335 does not allow data to be transferred according to the tagged storage location, invalidates or resets the tag 332 and skips to transfer the channel at the next storage location. This is to ensure that the circuit architecture of the present invention does not send out more data than it has taken in on the same channel. still,
It should be noted that a "skipover" can occur when there is not enough room to write data to the dual port memory 320, or when the bus is unavailable for transfer.

When using a proprietary circuit architecture to support channel interleaving, channel interleaving control structure 310 monitors the channel interleaving size, total transfer size and current transfer size per channel in both the write and read directions. By
Let the channel interleave the transfer. The channel interleave size for each channel is first for different channels.
Specifies the number of bytes that the first channel must transfer before it can interleave the transfer using the same resources as the channel. The total transfer size specifies the total number of data bytes that a particular channel has requested to transfer through the dual port memory 320.
The current transfer count is per channel for data bytes buffered in dual port memory 320.
Record both the current write count and the current read count that are in progress. In response to a request for data of a certain total transfer size over a channel, the channel interleave control structure 310 reads the channel interleave size of the channel and the total transfer size of the data slice that can divide the current transfer. Determine the number. Channel interleave control structure 310 if the data transfer can be divided into more than one data slice.
Can cause the second channel to perform the transfer after the first channel has transferred to its channel interleave size. The channel interleave control structure is written so that the circuit architecture of the present invention keeps track of the number of data bytes that have been transferred and the number of data bytes that are left untransferred for each channel, along with when channel interleaving can occur. Further monitor the current transfer count for each channel in both direction and read direction.
Therefore, the circuit architecture of the present invention is compatible with various DMs.
The transfer can proceed by performing data slice transfers from different channels until the A transfer sequence is complete. It is the channel sequencer 3 that orders the channels associated with the data slices.
30 and the channel interleave control structure 310.

Since various transfer sizes are associated with the channel interleave, the channel sequencer 330
A situation arises in which a channel transfer at one memory location of the memory is completed before a channel transfer at another memory location. This situation causes the channel sequencer 330 to initially
Despite the sequential loading of channels via their source pointers, they will have randomly distributed gaps. When an incoming channel is to be allocated to randomly distributed memory locations of the channel sequencer 330 to direct the transfer source, the input channel is between the source and destination pointers (ie, after the source pointer, It should be obvious to a person skilled in the art not to allocate to the memory location (before the destination pointer). Therefore, the new channel is not in the storage location where the destination pointer 335 would hit first, and the sequence of storage locations in the channel sequencer 330 is maintained.

Further, the channel interleave control structure 3
By comparing the channel interleave size with the difference between the total transfer size and the current transfer count, the data 10 has a size smaller than the channel interleave size of the channel, such as the remaining part of one transfer (“remnant”). Is to be transferred. When data of a size smaller than the channel interleave size is to be transferred through one channel, the channel interleave control structure 310 immediately after transfer of the first channel is completed so that the bus can be used more effectively. Let another channel perform the transfer. Also, when there is no other channel requesting data transfer after the first channel has completed the transfer, the channel interleave control structure 310 either continues the transfer on the first channel or responds to another request. Either the situation will be. Logically, the channel interleave control structure 310 may be a processor that processes information about the channel interleave size, the current transfer count and the total transfer size for each channel.

With reference to FIG. 4, the operation of the unique circuit architecture in the channel interleaving will be described with an example of the transfer operation. First, the DMA transfer is 32
Request 1000 bytes of data over channel 1 previously programmed to have a channel interleave size of 500 bytes to a host that supports bursts up to bytes. Note that the channel interleave size represents the maximum amount of data that can be transferred before another channel can be transferred over the same bus. The transfer begins with the memory 400 via the A bus 460 towards the dual port memory 420 and from the dual port memory 420 via the B bus 470 to the peripheral device 450 which requested the data. Channel interleave control structure 41
0 is how to “slice” a transfer by reading the channel interleave size, which is 500 bytes in this case associated with channel 1, the total transfer count, which is 1000 bytes, and the burst size, which is 32 bytes. Confirm whether to do. So, 15 individual 32 times
A 16-byte burst and a 4-byte transfer are added to the byte burst to write a 500-byte slice to the dual port memory 420, and the channel sequencer 430 sets "channel 1" to the slice of this data. Write as the corresponding source channel.

After channel 1 has written the maximum possible (500 bytes) slice of data without violating its 500 byte channel interleave size, another channel can use A-bus 460 to transfer the data. Like For example, if a DMA transfer via channel 7
If the use of bus 460 is allowed, then transfers via channel 7 can be interleaved at this point.
It should be noted that the computer system has multiple buffer architectures (not shown in FIG. 4) to accommodate DMA transfers so that another DMA channel can require the use of different buffer architectures for the transfer. May be. However, in order to clearly explain the operation of the present invention, it is assumed here that all channels use the same circuit architecture 401 to perform DMA transfers. If channel 7 is granted A-bus 460, and then channel 1's request to A-bus 460 is also granted, then transfers involving channel 1 are suspended until channel 7 reaches its corresponding channel interleave size. To do. The channel interleave control structure 410 causes the transfer parameters associated with channel 1, ie, the channel interleave size, to be suspended so that channel 1 can resume the transfer without being affected by the interleave of channel 7. Continue to maintain total transfer size and current transfer count.

The transfer of data over channel 7 is programmed so that channel 7 has different channel interleave size, total transfer size and burst size such that the size of the data "slice" associated with channel 7 is different from channel 1. The procedure proceeds the same as for channel 1, except that it is acceptable. Channel 7 slices from memory 400 to dual port memory 42
When written to 0, the channel sequencer 430 writes "channel 7" as the source channel corresponding to the slice of this data written to the dual port memory 420. If the next slice in the transfer is also from channel 7, then channel 7 is again written to channel sequencer 430 as the source channel.

After channel 7 completes the transfer from memory 400 to dual-port memory 420, or at least partially completes the data transfer allowed by the size of the channel interleave size, A-bus 460 grants the request. It is ready for use on any of the channels. Assuming channel 1 was granted the request,
Channel 1 is memory 400 to dual port memory 4
The data transfer to 20 can be resumed. Channel sequencer 430 writes "channel 1" as the source channel while writing another slice of data to dual port memory 420. FIG. 6 shows the current state of dual port memory 620 and channel sequencer 630. At the time of startup, the write pointer 625 and the read pointer 621 initially have the dual port memory 620.
Note that the source channel pointer 631 and the destination channel pointer 635 are also aligned with the first memory location of the channel sequencer 630.

Description will be made with reference to FIG. Channel 1
Continue to transfer the requested total transfer count by writing a second slice of 500 bytes to the memory location of dual port memory 620, starting at memory location 689, and then another channel count after the channel interleave size of channel 1 is reached. If the channel is allowed to use the bus, it suspends the transfer. If no interleaving occurs, channel 1 writes the remaining data to dual port memory 620, while channel sequencer 630 writes in parallel for each slice with "channel 1" as the source channel.

As shown in FIG. 6, a read pointer 621.
Is coupled to dual port memory 620. It should be noted that read pointer 621 and write pointer 625 operate independently of each other so that dual port memory 620 can be simultaneously writing and reading via its dual write / read pointer mechanism. When the data in the dual port memory 620 is to be read, the read pointer 621 indicates the dual port memory 62.
Read the data as it moves along 0. It is the destination channel pointer 635 that is coupled to the channel sequencer 630 that points to the destination of the data. Thus, data is read from dual port memory 620 via its destination channel before violating the channel interleave size of that channel. Destination channel pointer 635
Increments to the next memory location in the channel sequencer 630,
It points to the destination channel of the data to be read from the dual port memory 620 via the read pointer 621. The read operations of dual port memory 620 can also be interleaved to support multiple channels, as described for the write operations of dual port memory 620. Further, if the total transfer count is reached when reading data from dual port memory 620, another channel is allowed to read data from dual port memory 620 even though the remaining transfers are less than its channel interleave size. Therefore, if there is no more data to be transferred after transferring the remnant, the other channel is not forced to wait and the bus can be used more efficiently. Currently, the dual port memory 620 is realized by two dual port random access memories as shown in FIG.

Although the circuit architecture of the present invention has been described as the preferred embodiment for supporting channel interleaved DMA operation, it should be apparent to those skilled in the art that the circuit architecture can be readily applied to other multiple device transfer operations. Is. For example, in the case of data transfer between a CPU and multiple devices via a common transfer resource, the circuit architecture can maintain the transfer sequence of each device while performing individual read and write operations in succession. . With such a circuit architecture, transfer resources can be efficiently used. During operation, the device-specific data is stored in the dual-port memory while the device ID is recorded in the sequencer. Similar to the channel interleaving mechanism, a data transfer count can be maintained for each device to monitor the status of the data stream. Compared to the "one-buffer-per-device" approach, the circuit architecture does not sacrifice the silicon area required by all of the individual buffers, allowing increased bandwidth and improved operational continuity. .

[Brief description of drawings]

FIG. 1 is a schematic diagram of a DMA controller module.

FIG. 2 is a block diagram of a computer system with a DMA controller module.

FIG. 3 is a schematic diagram of a circuit architecture supporting multiple interleaved DMA channels incorporating the teachings of the present invention.

FIG. 4 is a schematic diagram of a circuit architecture incorporating the teachings of the present invention.

FIG. 5 is a schematic diagram of circuit architecture in a preferred embodiment.

FIG. 6 is a schematic diagram of circuit architecture in an example of a write operation.

FIG. 7 is a flowchart showing a channel interleaving process.

[Explanation of symbols]

 200 system bus 210 CPU 220 DMA controller module 230 buffer 240 memory 250 peripheral device 300 control structure 310 channel interleave structure 320 dual port memory 321 read pointer 325 write pointer 330 channel sequencer 331 source pointer 335 destination pointer 400 memory 401 circuit architecture 410 channel Interleave control structure 420 Dual port memory 430 Channel sequencer 450 Peripheral device 460 A bus 470 B bus 620 Dual port memory 621 Read pointer 625 Write pointer 630 Channel sequencer 631 Source channel pointer 635 Destination channel pointer 689 Storage location

Claims (3)

[Claims] 1. A method of transferring data using a transfer resource by a plurality of DMA channels each having a predetermined channel interleave size, wherein the transfer resource is used to transfer a first number of data. A first DMA channel for providing a first DMA channel for comparing the first number with a first predetermined channel interleave size of the first DMA channel; Transferring the first number of data over the first channel using the transfer resource if the channel interleave size is less than or equal to a predetermined channel interleaving size; Dividing into a plurality of slices having a size not exceeding one predetermined channel interleave size; Transferring a slice using the transfer resource; for determining whether a second DMA channel has obtained control of the transfer resource after each slice of data in the first DMA channel Inquiring, if the second DMA channel is not in control, continuing to transfer slices of data using the transfer resources by the first DMA channel; Transferring data through the transfer resource and transferring control of the transfer resource to the second DMA channel so as to interrogate after each slice. 2. An apparatus for interleaving data transfer using one transfer resource by a plurality of DMA channels each having a predetermined channel interleave size, the data being coupled to said transfer resource and transferred for a first channel. The total transfer count, which is the total number of
Receiving means for receiving from said channel; control means coupled to said receiving means for determining whether said total transfer count is greater than said predetermined channel interleave size of said first channel; coupled to said control means If the total transfer count is greater than the predetermined channel interleave size, the data from the first channel is divided into a plurality of slices, each slice being less than or equal to the predetermined channel interleave size for the first channel. Data slicing means for dividing into a plurality of slices; transfer means for transferring the plurality of slices by the first channel; Keep the current transfer count, which is the number of data already transferred A first counter means that, the first
1 from said first channel coupled to a counter means of
After one slice has been transferred, it is determined whether the second channel has control over the transfer resource, and if so, the interrogation means for coupling the transfer means to the second channel. Coupling to the transfer means and determining whether the total transfer count for the first channel is equal to the current transfer count for the first channel, and if so, all data has already been transferred. Second counter means for coupling the transfer resource to the second channel so that sometimes the second channel does not have to wait for the first channel to terminate the channel interleave size of its data; A device comprising. 3. A plurality of DMA channels each having a predetermined channel interleave size, the first DMA channel of which waits to transfer a first number of data via transfer resources. In an apparatus for transferring data using a transfer resource, the first number is compared with a first predetermined channel interleave size of the first DMA channel to determine that the first number is the first number. Is determined to be equal to or smaller than a predetermined channel interleave size of the first channel, and is equal to or smaller than the first predetermined channel interleave size, the first number of data is transferred to the first channel Comparing means for transferring the first number of data over a plurality of slices each of which does not exceed the first predetermined channel interleave size. Data dividing means for dividing into data;
Data transfer means for transferring each slice of data by the first DMA channel using the transfer resource; and after transferring each slice of data by the first DMA channel, the second DMA channel Inquiring means for determining whether or not control of the transfer resource has been obtained, and if not, inquiring means for continuing to transfer a slice of data using the transfer resource by the first DMA channel; Control channel releasing control of said transfer resource to said second DMA channel so that said DMA channel transfers its data via said transfer resource and interrogates after each slice.